Distribute model volume to nearest tube cell midpoint

[wouterjdb]

This PR changes the way volume from an original simulation model is distributed to FlowNet. Instead of distributing to the mid point of all tubes the volume is distributed to the midpoint of all tube cell mid points.

The volumes distributed to flownet tube cells are then summed per tube, and evenly redistributed over the single tube.

---

Contributor checklist

• 🎉 This PR closes Too much volume assigned to (short?) tubes #395.
• 📜 I have broken down my PR into the following tasks:
  • Change algorithm
  • Test
• 🤖 I have added tests, or extended existing tests, to cover any new features or bugs fixed in this PR.
• 📖 I have considered adding a new entry in CHANGELOG.md.
• 📚 I have considered updating the documentation.

[wouterjdb]

Comparison of volumes (identical color scales):

Old method

New method

Observations:
The total distributed volume is similar (6.3e8 vs 6.5e8); minor differences are to be expected due to differences is porosity and bulk volume multipliers being different. Using the new distribution scheme there are no extreme outliers anymore for short vertical connections.

The following plot shows the volume distribution using the new algorithm, but now with a color scale that matches better with the assigned volumes.

Some outliers can still be observed, but the outlier that has the highest volume per cell has less than 40% of the volume that was previously assigned to the largest grid cell - and there are fewer outliers. This suggest that all-in-all the volume is more evenly distributed (as in, a better approximation of the original model).

Looking at the statistics, the new method has a similar P10, mean and P90 values for the distribution (but keep in mind that number of cells are not 100% identical). However, again, the maximum values has significantly reduces - the outliers are reduced.

[olelod]

LGTM